Fiber Optic Communication Link Design

By

Michael J. Fujita, S.K. Ramesh, PhD, Russell L. Tatro

Abstract – The fundamental building blocks of an optical fiber transmission link are the optical source,
the optical fiber transmission medium with associated connectors, and the photodetector with its
associated receiver [1]. The system designer must select from a set of device components to meet a given
set of system requirements. Analysis and/or simulation are required to validate that the designed system
meets those requirements.

EEE 165, CSU Sacramento

December 1, 2003
TABLE OF CONTENTS

Page
List of Tables i
List of Figures i
I. Introduction 1
II. Components 3
III. System Requirements 4
IV. Top Down Design 8
V. Conclusion 8
References 9

LIST OF TABLES

Table Page Caption

I 6 System Power Budget

LIST OF FIGURES

Figure Page Caption

1 1
2 2
3 2
4 4
5 5
6 6
7 7
Link Diagram.
Eye diagram.
Eye diagram interpretation.
Component loss diagram.
Receiver sensitivities.
Graph of power budget.
RZ and NRZ coding schemes.

Page i
 I. INTRODUCTION

A fiber-optic communication system comprises three basic elements: a compact light source, a low-loss and
low-dispersion optical fiber, and a photodetector [2]. We have already covered these devices individually
earlier in the course. In this handout, we examine their role in the context of the overall design, operation and
performance of an optical communication link. Optical accessories such as connectors, couplers, switches,
splices and multiplexing devices are essential to the successful operation of fiber links and networks [2].
From a selection set of device components and a given set of system requirements, analysis and/or simulation
is required to determine if the system being designed will meet the requirements [3].

Figure 1 - Link Diagram [4].

As a first approximation one can employ an analytical method for link power budget to determine power
attenuation limits and system rise time budget to verify data integrity. The maximum fiber length that can be
used to transmit data at a given data rate and with a prescribed level of performance is determined. The
analytical method relies on static, worst-case performance values of the components, and will usually result
in an over-designed and higher-cost system.
Computer-aided modeling and simulation (such as LinkSim by Rsoft, or MATLAB SIMULINK) will
allow device operation around an average value, and rely on statistical methods in the design approach. In
addition, as link systems become large and complex, device parameter modeling, noise source modeling, and
measurement methods of waveform distortion such as the eye-diagram are essential for characterizing the
performance. Often, an iteration of experimental verification of particular elements into the device modeling
is essential to build valid and reliable state-of-the-art system simulation models. Whenever possible a
correlation of the model with experimental results should be performed. This will provide validation of
modeling techniques and confidence regarding its use for alternate parameter selections.
For digital systems, the accepted performance criteria is a bit error rate of BER ≤ 10-9. This is a bit error
rate where no more than one error should occur in the transmitting of 109 data bits. In analog terms this is a
signal to noise ratio SNR = 21.6 dB or in linear terms a SNR = 144.

 Signal 
 
BER = ½ erfc  Noise 
(1)
 2 2 
 
 

where erfc is the error function.

Page 1
 In the lab, a test called the eye pattern measures the system performance. The test is named for the
appearance of the signal on an oscilloscope. As the figure below shows, the pattern resembles an eye.

Figure 2 - Eye Diagram [4].

The measurements are made in the time domain and allow the effects of waveform distortion to be
shown immediately on an oscilloscope. A random data signal generates ones and zeros at a uniform rate
but in a random manner. When this random data stream is the input signal, then the eye pattern is
presented. Figure 2 a) shows an ideal response to this random pattern. However, in a real system,
distortion of the pulse, and nonlinearities result in a pattern similar to figure 2 b).

Figure 3 - Eye diagram interpretation [1].

The eye pattern reveals the following (in brief):

Width of the eye – maximum time interval without intersymbol interference.
Height of the eye – shows the noise margin. The ratio of peak signal voltage (V1 in Figure 3) to the
maximum signal voltage (V2 in Figure 3) is the noise margin which is a measure of the immunity to
noise of the system.
System rise time (10% to 90% of max value) is easily measured.
Nonlinearities in the system will result in an asymmetry in the eye pattern.

Page 2
 II. COMPONENTS

A. Photodetector

The photodetector is typically a pin or avalanche photodector (APD) device. Thermal detectors do not
achieve a bandwidth useful for communications. Whatever the specific photodetector, you must examine
the responsivity R at the operating (source) wavelength. Can the device create a sufficient photogenerated
current at the source wavelength? This depends on the sensitivity of the device. That is, it’s noise
characteristics must meet the required signal to noise ratio (or carrier to noise ratio) at the desired
bandwidth.

I photocurrent I ph
1) Responsivity R = = (2)
Powerincident Po
2) Speed – usually given as bandwidth in Hz
3) Operating wavelength - set by light source
4) Sensitivity – What is the minimum incident light power that can be detected by the photodetector and
receiver combination reliably.

B. Optical Source

The basic requirements for the lights sources used in optical communication systems depend on the
nature of the intended application (long-haul communication, local-area network, etc.). [2] Both light-
emitting diodes (LEDs) and laser diodes are used as sources. Laser diodes have the advantage of high
power (tens of mW), high speeds, and narrow spectral width. LEDs have the advantages of ruggedness,
reliability, lower cost, long lifetime, and simplicity of design. However, they suffer from relatively
broader linewidth and lower bandwidth.
Only that light which can couple to an optical fiber is useful. The output light angle is given by

NumericalAperture
α max = sin −1 ( ) (3)
η0
where η0 is the interfaces refractive index.

Thus the size of the emitting area and the emitting pattern can be important to the amount of light coupled
to the fiber. The source will also have a finite spectral wavelength ∆λ around the center emission
wavelength. This ∆λ can give rise to dispersion if coupled to the fiber. If we use external wavelength
filters, this may decrease the power coupled. Even at a single “color” (wavelength) several modes may be
present. Coupling to a single-mode fiber will lose the power in these other modes.
Thus the main features for a light source are:
1) Power. The source power must be sufficiently high so that after transmission through the fiber the
received signal is detectable with the required accuracy. This includes stability of that power both in
amplitude and frequency.
2) Speed. It must be possible to modulate the source power at the desired rate.
3) Spectral linewidth. The source must have a narrow spectral linewidth so that the effect of chromatic
dispersion in the fiber is minimized. What is the effect of a finite ∆λ (dispersion) on your system?
4) Noise. The source must be free of random fluctuations.
5) Emission wavelength. Contrast to fiber attenuation.
6) Emission pattern – The source power must be coupled to the fiber. Both light output area and mode
distribution in the beam can be important.

Page 3
C. Optical Fiber

Light waves travel in the optical fiber in the form of modes, each with a distinct spatial distribution,
polarization, propagation constant, group velocity, and attenuation coefficient. Thus the fiber chosen for
your system must meet the design goals after consideration of:

1) Multi-mode fiber whether step-index, or graded-index. Single-mode fiber.

2) Core size
3) Core refractive index profile
4) Bandwidth and dispersion characteristics
5) Attenuation dB/km (including any connectors and splices) which is a function of central wavelength.
6) Numerical aperture NA = n12 − n22 (4)

III. SYSTEM REQUIREMENTS

A. System Overview

The design of an optical link involves many interrelated variables among the fiber, source and
photodetector operating characteristics. In carrying out an optical fiber link analysis, several iterations
with different device characteristics may be required before it is satisfactorily completed. The key
requirements needed in analyzing a link are: [1]

1) Desired (or possible) transmission distance

2) Data rate or channel bandwidth
3) Bit error rate (BER) allowed. Signal to noise ratio (SNR) allowed for an analog system

Figure 4 - Component loss diagram [1].

Page 4
B. Power Budget

Power budget considers the total allowed optical power loss PT between the source and the detector.
The designer must account for losses due to fiber optic fiber attenuation, connectors, splices and allow for
a “system margin” as a safety factor. The measure of the total power loss is given as:

Pout
Ploss = 10 Log10 ( ) (5)
Pin

Suppose the designer has the following requirements.

Data rate = 20 Mb/s
Bit error rate = 1 x 10-9 errors/bit
A “short haul” system with an operating wavelength λ = 850 nm.

A Si p-i-n photodiode is selected as the photodetector. What is the receiver sensitivity at this wavelength?

Figure 5 - Receiver sensitivities [1].

The above graph shows that the Si pin photodiode has a receiver sensitivity of about –42 dBm (63.1 nW)
at the required bandwidth of 20 Mb/s and in the wavelength range 800 – 900 nm.
Now suppose you chose to use a GaAlAs LED for this short haul system. The data sheet for the LED
states that it can achieve Pcoupled = 50 µW into a multi-mode fiber with a core diameter of 50 µm. The
power includes all coupling loss involved in launching the power into the fiber. A point-to-point system
(no repeaters) will be used, so there is one connector to the detector. The loss is Lc = 1 db per connector.
Splices will be required for each kilometer of fiber. The attenuation loss for the fiber which was
previously installed is αF = 3.5 dB/km. You decide to allow for a sufficient safety margin for temperature
ranges, ageing effects and future splices. The following table shows the Power Budget.

Page 5
 Table I. - System Power Budget.
Optical Transmitter Output Power -13 dBm
Receiver sensitivity -42 dBm

Required margin 29 dBm

System Loss:
Fiber Loss 3.5 dB/km X 5 km 17.5 dB
Connector Loss 1dB x 2 connectors 2 dB
Total Splicing Loss (0.5 dB X 5) 2.5 dB
System margin (Headroom) 6 dB

Total Attenuation 28 dB

So in this example, we have an excess power margin of 29 dB – 28 dB = 1 dB. Thus we have a total
headroom of 7 dB when we include the system margin of 6 dB. The following figure graphically presents
this information.

Figure 6 - Graph of power budget [1].

C. Rise Time Budget

The rise time budget determines the dispersion limitation of a link. Dispersion is the broadening of a
pulse as it travels through an optical fiber as discussed earlier in the course. We will define a rise time
due to the material (chromatic) (tmat) and modal dispersion (tmod). Thus the total rise time of the system is
the rms sum of all the rise times.

N
Total rise time: tsys = ∑t
i =1
i
2
= t tx2 + t mat
2
+ t mod
2
+ t rx2 (6)

Page 6
For digital data, two main coding schemes are used. They are the return-to-zero (RZ) and non-return-to-
zero (NRZ). The NRZ method occupies the entire bit width (bit period), whereas the RZ method occupies
only a portion of the bit period.

Figure 7 - RZ and NRZ coding schemes [4].

A general rule of thumb is the system rise time should be less than a percentage of one of these methods.
The bit period in seconds is the reciprocal of the bit rate in Hertz.
 1 
tsys(RZ) ≤ 35%   = 35% (Bit Period) (7)
 BitRate 

 1 
tsys(NRZ) ≤ 70%   = 70% (Bit Period) (8)
 BitRate 

The rise times of the optical transmitter and receiver are usually know to a designer. Define ttx as the light
source and drive circuit rise time. The optical detector and receiver circuitry rise time can be
approximated by trx(in nsec) = 350/Brx where Brx in MHz is the rms of the photodetector response
bandwidth and the 3 dB electrical bandwidth of the receiver circuitry. [1]

The rise time due to material (chromatic) dispersion is given by

tmat = Dmat σλ L (9)

Where σλ is the spectral width of the optical source. This is a little bit of nomenclature variation here. In
the Kasap text, the value is given as ∆λ1,2 which is full width at half maximum value. The Dmat term is the
material dispersion factor of the fiber to be used and is usually given in ns / (nm x km).

The rise time due to intermodal dispersion in the optical fiber can also be approximated if necessary.
The bandwidth of a fiber of length L is defined as BM(L). B0 is the bandwidth distance product for a 1 km
length of fiber as given by the product’s specifications. The power q of L in the equation below has been
found thru experiment to be between 0.5 and 1. Here we will let q = 0.7 as a reasonable estimate for
design. Thus the bandwidth distance product over length L can be estimated as [1]:

B0
BM(L) = (10)
Lq

Page 7
 IV. TOP DOWN DESIGN ANALYSIS [5]

Step 1 – What wavelength do we wish to operate at?

- “Short haul” (e.g. LAN): 800 – 900 nm
- “Long haul” (e.g. transatlantic cable): 1300 or 1550 µm

Step 2 – Select a photodetector

- What is minimum light power incident on the detector as driven by the BER or SNR?
- Should I pick a pin or an APD type of detector?

Step 3 – Select an optical light source

- Criteria includes Cost, signal dispersion, and distance
- 800 – 900 nm LED: 150 (Mb/s) (km) data-rate distance product
Laser: 2.5 (Gb/s) (km)
- 1300 to 1550 nm LED: 1.5 (Gb/s) (km)
Laser: 25 (Gb/s) (km)

Step 4 – Select an optical fiber

- LEDs usually connect to multi-mode optical fibers in order to launch an acceptable amount of
light power
- Lasers can connect to either multi-mode or single-mode fibers. What are the
advantages/disadvantages of selecting a single-mode fiber?

V. CONCLUSION

An optical link design must meet both the link power budget and the system rise time analysis. In the
link power budget analysis one first determines the power margin between the optical transmitter output
and the minimum receiver sensitivity needed to establish a specified BER (or SNR for an analog system).
Once the link power budget has been established, the designer makes a system rise time analysis to ensure
that the dispersion limit of the link has not been exceeded [1].
In designing an optical-fiber system, one should take into account that the individual component
parameters can vary considerably. Statistical analysis can refine the design to a less conservative
approach at the assumed cost of a small number of failures. In the future such an analysis will become
increasingly important as size, complexity and speed of links continue to evolve.

Page 8
 REFERENCES

[1] G. Keiser, Optical Fiber Communications, 2nd ed., New York, McGraw Hill , 1991.

[2] B. Saleh and M. Teich, Fundamental of Photonics, New York, Wiley & Sons, 1991.

[3] H. Tanzer, Course materials for EEE 167, California State University Sacramento, 2002.

[4] J. Wilson and J. Hawkes, Optoelectronics, An introduction, 3rd ed., London, Prentice Hall Europe,
1998.

[5] M. Fujita, Course materials for EEE 165, California State University Sacramento, 2001.

Page 9
Problem 9.8 [4] An emitter can couple some 10-3 W of optical radiation at 900 nm into a fiber which has
an attenuation of α = -5 db/km. A receiver is used which requires an average of 500 photons per bit for
an acceptable bit error rate. Plot a graph of the maximum fiber length possible for bit rates of between
105 and 109 bits per second (bps). Modify this diagram to include the effects of fiber dispersion of 5
ns/km if the maximum allowable pulse spreading must not exceed 1/(2 x bit rate).

Solution:

If Pout is the power launched into the optical fiber and Pin is the minimum detectable power at the detector,
then the total loss thru a length of fiber is given as:

Pout
Loss = 10 log10 where the loss = attenuation x distance = α L
Pin

Thus the distance L is then:

Pout
10 log10 ( )
Pin
L=
α

The minimum power (PD) that can be detected is given as

PD = (number of photons/bit)(Bits/sec)(Energy/photon)
500 photons Bits hc
= X X
bit sec λ
500 photons Bits  Joules 
= X X  2.21x10 −19 
bit sec  Photon 

PD(105 bps) = 1.104 x 10-11 W

PD(106 bps) = 1.104 x 10-10 W
PD(107 bps) = 1.104 x 10-9 W
PD(108 bps) = 1.104 x 10-8 W
PD(109 bps) = 1.104 x 10-7 W

Now we can find the distance L for each of these detectable powers PD.

L(105 bps) = 15.9 km

L(106 bps) = 13.9 km
L(107 bps) = 11.9 km
L(108 bps) = 9.9 km
L(109 bps) = 7.9 km

You’ll see this attenuation limited fiber length plotted on a graph shortly. First, let us look at the effects
of dispersion which results in pulse broadening.

Page 1
The problem gave the maximum allowable pulse spreading τ must be less than 1/(2 x bit rate). The
dispersion per kilometer σ = 5 ns/km. Thus τ = σ L ≤ 1/(2B). In terms of length, we can write

1
L=
σ 2B

L(105 bps) = 1000 km

L(106 bps) = 100 km
L(107 bps) = 10 km
L(108 bps) = 1 km
L(109 bps) = 0.1 km

Now look at the graph on the next page. Two lines are plotted. For a given bit rate, when one looks up
from the horizontal axis, the first line you run into is the limiting factor. Thus, above 107 bit per second,
dispersion limits the fiber length. Below 107 bit per second, attenuation limits the fiber length.

Page 2
Page 3
Problem 9.14 [4] It is required to construct an optical fiber communications link 5 km long with an
analog modulation bandwidth of 10 MHz. The proposed system uses multimode step index fiber with
core and cladding refractive indices of 1.48 and 1.46 respectively. At the operating wavelength the fiber
exhibits an attenuation of 3 db/km. The emitter is an LED which can launch 200 µW of optical power
into the fiber and the detector is a p-I-n photodiode which requires –40 dBm to achieve the required
signal-to-noise ratio (SNR). No splices are required. The detector and emitter have rise times of 100 ns
and 10 ns respectively. Determine whether the link should operate satisfactorily and if not discuss the
component changes you would recommend.

L = 5 km
Bandwidth (BW) = 10 MHz
Fiber attenuation α = 3 dB/km at central wavelength λ0
trx = 100 ns
ttx = 10 ns
Pmin at the detector = -40 dBm = 100 nW

Solution:

Power Budget
The power launched into the fiber PLED = 200 µW = 10 log10 (200 µW/1mW) = -7 dBm

Total power budget PT = PLED – Pmin(at the detector) = -7 dBm – (-40 dBm) = 33 dBm

System Losses
Fiber Loss 5 km x 3 dB/km 15 dB
Detector – assume one connector at 1 dB 1 dB
Splices or other connectors – none 0 dB
System Margin (Headroom) 6 dB

Total Losses 22 dB

So the excess power margin is 33dB – 22 dB = 11 dB which seems very ample in this case.

Now for the sake of argument, let us suppose that the LED is capable of emitting 200 µW of power, but it
must still be coupled to a fiber. The coupling efficiency ηc is given by:

NA 2 n12 − n22 1.48 2 − 1.46 2

ηc = 2 = = = 0.242
n0 n02 12

where the interface between the fiber and the LED is assumed to be air with n0 = 1.
Thus the power coupled by the LED into the fiber is

PLED x ηc = 200 µW x 0.242 = 48.4 µW = -13.2 dBm.

The system margin now becomes –13.2 dBm – (-40 dBm) = 26.8 dBm. The excess power margin is still
4.8 dBm, and the system power budget is acceptable.

Page 4
Rise Time Budget
The bandwidth BW = 10 MHz. So the pulse period is 1/BW = 100 nsec. Also note that the problem did
not give a central wavelength nor a spectral linewidth ∆λ. Thus we have insufficient data to determine
the dispersion. So the total system rise time with dispersion excluded is:

tsys = t tx2 + t rx2 = 100ns 2 + 10ns 2 = 100.5 ns

Since tsys > allowed pulse width (100 ns), the systems design fails even without including the effects of
dispersion. The component that must be changed is the detector. A faster photodetector is needed. But
be sure to account for the total power budget with the new photodetector’s minimum power sensitivity.

Page 5